[[
wikihub
]]
Search
⌘K
Explore
People
For Agents
Sign in
Explore
People
For Agents
Sign in
@jemoka / Jemoka Knowledge Base / raw/concept/kbhmultiple_instance_learning.md
Suggest edit
Cancel
Submit suggestion
Title
Name
Note
--- title: "Multiple Instance Learning" source: https://www.jemoka.com/posts/kbhmultiple_instance_learning/ --- \begin{equation} B = \qty[(x_1, y_1), \dots, (x_{n}, y_{n})] \end{equation} where the labels would be: \begin{equation} C(b) = \begin{cases} 0, if \sum_{i}^{}y_{i} = 0 \\ 1, \text{otherwise} \end{cases} \end{equation} and then we maxpool MILFormer MILFormer is a multiple-instance learning scheme which makes predictions over input patches whose output predictions are weighted as multi-distirbution.